How to design a learning experience with AI

Define the performance objective

Specify the learner, task, conditions, quality criteria, and boundary. “Understand verification” is not observable. “Given an AI-assisted brief and its cited sources, identify unsupported claims, correct the brief, and document remaining uncertainty” can guide practice and assessment.

CDC Quality Training Standards connect needs assessment, learning objectives, relevant content, engagement, evaluation, learner assessment, accessibility, and follow-up support. OPM’s instructional-systems guidance likewise connects occupational analysis, objectives, strategy, performance measures, and validation.

Map the authentic task

Break the target performance into decisions and actions. Identify prerequisite knowledge, common errors, tools, time pressure, and evidence available in real work. Preserve the assessment-product boundary: a learning simulation can create an opportunity to observe task-bounded behavior, not prove performance everywhere.

Worked synthetic design

Audience: analysts beginning to use AI for decision briefs.

Objective: Within 35 minutes, produce a one-page recommendation that distinguishes evidence, inference, and unknowns; verifies three material citations; and escalates one unresolved conflict.

| Learning event | Learner action | Feedback | Evidence | | --- | --- | --- | --- | | Contradictory source pair | Compare authority and date | Facilitator models source map | Completed map | | AI draft with citation traps | Inspect and correct claims | Immediate rationale feedback | Error log | | Timed scenario | Deliver revised brief | Rubric plus debrief | Brief and workflow record | | Workplace follow-up | Apply checklist to approved task | Manager discussion | Reflection and artifact |

All tasks and evidence are synthetic.

Sequence practice, not slides

Begin with a worked example, then supported practice, contrasting cases, and a realistic whole task. Fade prompts as capability develops. Include errors that resemble real failures rather than trivia. Give feedback on the reasoning and verification process, not only the final answer.

Ask AI to create variations that change source authority, missing data, or time pressure. Review every generated fact and citation before use.

Align assessment with the objective

If the objective is source verification, a recall quiz is insufficient on its own. Use an artifact or observed task with the same decisions, tools, and constraints. Define the rubric and missing-evidence treatment before delivery. Provide accommodations and accessible alternatives that preserve the construct.

Plan transfer support

Add job aids, manager prompts, peer review, follow-up practice, and access to approved tools. Distinguish learning evidence from workplace transfer. Course satisfaction and completion do not establish application.

Verification checklist

Confirm the objective is observable; tasks reflect real work; content is necessary; practice increases in complexity; feedback is actionable; assessment matches conditions; sources are verified; accessibility is reviewed; follow-up support exists; and outcome claims stay within collected evidence.

Pilot the experience with representatives of the intended audience. Record where instructions, tools, timing, prior knowledge, or accessibility barriers obstruct performance, and revise those conditions separately from the assessment criteria. A learner struggling with a broken source link has not demonstrated the targeted capability poorly. Keep pilot observations, changes, and reasons together so later reviewers can inspect design decisions. Recheck the revised experience with a fresh task instance and record remaining friction.

Sources and claim map

The audience, scenario, timing, and outcomes are synthetic. This educational design does not claim validated learning, transfer, or assessment results.

• Centers for Disease Control and Prevention, Quality Training Standards: supports needs-led design, objectives, relevant content, engagement, assessment, accessibility, evaluation, and follow-up support.
• U.S. Office of Personnel Management, Instructional Systems Series 1750: supports linking learner characteristics, objectives, instructional strategy, performance measures, and validation.

Sources: [1] [2]